Where to Start with Automation and AI: A Practical Guide

Many organisations recognise the potential of automation and artificial intelligence. However, deciding where to begin is often the hardest step. Leaders may know that manual work exists across the business but lack clear evidence about where automation would deliver the greatest value.

Research suggests the opportunity is significant. Workers who use automation save an average of 3.6 hours each week. Yet only about 45 percent of teams have implemented automations, despite 77 percent saying automation would improve productivity (Slack).

At the same time, knowledge workers spend substantial time on low value activities. Employees spend around 1.8 hours each day searching for information, according to research cited by Cottrill Research.

These numbers highlight the scale of inefficiency. However, organisations cannot prioritise automation effectively without understanding how work actually happens day to day.

A practical starting point is measurement. By gathering simple data about time spent, errors and delays, businesses can identify which processes offer the strongest automation opportunities.

Why measurement should come first

Automation initiatives often begin with discussions about technology. In practice, successful projects usually start with understanding the current process.

Without baseline data, leaders struggle to answer basic questions.

• how much time does this process consume?
• how often does it occur?
• where do delays or errors happen?
• what would improvement actually be worth?

Without answers to the above questions in a specific organisation, the scale of the problem (or potential saving) remains unclear.

Measurement provides the evidence needed to justify investment.

Measurement also improves compliance and risk management.

Manual processes introduce variability, and error rates often rise as workloads increase. Tracking metrics such as manual effort eliminated, error reduction and service level performance can also help quantify the impact of automation (Moveworks).

Metrics that reveal automation opportunities

Many automation opportunities become visible once a few simple metrics are tracked. These measurements do not require complex systems and can often be gathered using existing tools. For example:

Manual effort and time consumption

The first indicator is the amount of manual effort required to complete a task.

Key metrics include:

• hours spent per task,
• average time required to complete a transaction,
• number of employees involved in completing the work.

Processes that consume significant staff time are often strong automation candidates.

Frequency and transaction volume

Automation delivers greater value when processes occur frequently.

Useful metrics include:

• number of transactions per day or week,
• number of requests handled by a team,
• total workload across departments.

High volume processes with predictable steps often deliver the fastest returns when automated (RTS Labs).

Error rates and rework

Manual processes frequently create errors or require corrections.

Tracking these indicators can reveal hidden inefficiencies.

Such as:

• number of corrections required per transaction,
• percentage of work requiring rework,
• number of emails or handoffs needed to complete tasks.

High rework rates often indicate that a process lacks consistency. Automation can help enforce standard steps and reduce variation.

Cycle time and bottlenecks

Cycle time measures how long it takes for work to move from request to completion.

Long cycle times often reveal bottlenecks caused by manual approvals, information searches or repeated handoffs.

Even simple timestamp tracking can reveal delays between process steps. More advanced approaches, such as process mining, analyse system logs to reconstruct real process flows and identify bottlenecks (SAP Signavio).

Customer and employee experience

Automation decisions should also consider experience metrics.

Examples include:

• customer response and resolution times,
• employee wait time for approvals or information,
• satisfaction scores or feedback from internal teams.

Improvements in these areas can have direct business impact, particularly in customer facing processes.

Simple ways to gather useful data

Many organisations assume process measurement requires specialised tools. In practice, useful data can often be gathered using simple methods.

Manual time tracking

Employees can record how much time they spend on specific tasks for a few days.

Even short samples can provide valuable insights when extrapolated across a week or month.

Short surveys and interviews

Employees usually know where inefficiencies exist.

Brief surveys or interviews can reveal:

• time spent searching for information,
• tasks employees find repetitive or frustrating,
• processes that create delays.

These insights often highlight automation opportunities that leadership teams have not recognised.

Sample observations

Observing a small number of transactions from start to finish can reveal process complexity.

Tracking a few real examples helps estimate cycle time, error rates, and the number of manual steps involved. This approach is particularly useful when processes span multiple departments.

Using existing system data

Many business systems already capture useful process information.

Examples include:

• CRM systems recording ticket response times,
• ERP systems tracking invoice processing steps,
• help desk platforms capturing request volumes.

Extracting data from these systems can reveal bottlenecks without requiring new measurement tools.

Advanced approaches such as task mining can analyse user activity and generate detailed metrics such as average handle time and manual rework (Mimica).  However, these tools are usually most useful after simpler measurement has already revealed significant inefficiencies.

Building a business case for automation

Once measurement data exists, organisations can begin building a structured business case.

A practical approach includes the following steps.

Document current processes

Map how work actually occurs, including decision points, handoffs and delays.

This step often reveals hidden complexity that process owners were not aware of (Superhuman).

Identify repetitive tasks

Look for activities that follow predictable rules and require limited human judgement.

Common examples include:

• data entry,
• invoice processing,
• report generation,
• customer query handling.

These types of tasks are typically well suited to automation.

Estimate potential savings

Once manual effort is measured, organisations can estimate financial impact.

A simple calculation multiplies hours saved by the average hourly cost of the employees performing the work.

One example analysis showed that automating a process saved 257 hours per month, representing approximately £10,000 in labour cost (MaxHR).

Prioritise opportunities

Not every automation opportunity should be pursued immediately.

Many organisations use an impact versus effort framework.

This approach prioritises initiatives that deliver significant benefit while having relatively low implementation complexity (Superhuman).

Start small and scale

Automation programmes rarely succeed when launched as large transformation projects.

Instead, organisations typically begin with small pilot projects. A single process or department can serve as a test environment.

During the pilot phase, teams should track:

• time saved,
• error reduction,
• employee feedback,
• customer experience improvements.

These results provide evidence that automation delivers value.

Once benefits are proven, the same measurement approach can identify additional processes suitable for automation.

Over time, organisations can expand automation across departments while continuously refining their measurement methods. Monitoring metrics such as time saved, productivity improvements and error reduction helps ensure automation continues to deliver measurable business impact (Riseup Labs).

Takeaway

The starting point for automation and AI is not technology. It is understanding how work is currently performed.

By measuring manual effort, error rates, cycle times and process volumes, organisations can identify where inefficiencies exist and where automation will deliver the greatest value.

Importantly, this measurement does not necessarily require complex tools. Simple time tracking, employee feedback and existing system data can reveal substantial opportunities.

Once these insights are available, organisations can prioritise high impact processes, test automation through pilot projects and gradually expand successful solutions.